Hi,
Given A design matrix:
EV0 EV1 EV2 EV3
1 1 10 50
1 1 20 90
1 -1 10 200
1 -1 2 5
1 1 4 5
1 -1 30 8
first_utils will demean EV1,EV2,EV3. If the ones column is not in the design
matrix first_utils will demean the other EVs and add the ones column.
So, to test for an effect on each EV, the contrasts for EV1, EV2, EV3 would
be [0 1 0 0], [0 0 1 0], [0 0 0 1] respectively. first_utils by default
outputs the results for each contrast (specified by the 1,2,3 appended to
the file name). Currently, first_utils does not allow you to specify a
contrast.
For the surface output, the mean surfaces correspond to EV1 and are the same
for each contrast. The group membership from which the surface means are
evaluated are based on those less than zero versus those that are greater
than zero (for EV1 after demeaning).
Hopefully this clear thing up,
Cheers,
Brian
> sorry to keep piling up the questions, but if i set up my regression for
> first_utils as one group, with one regressor, then only try to look at
> that regressor (just like i would in feat), where does first_utils split
> them to create the 2 means displayed in the .vtk?
>
> ie:
>
> EV1 EV2
> 1 10
> 1 5
> 1 30
> 1 12
>
> contrast:
> 0 1
|